[Paper Review] Linear Stationary Iterative Methods for the Force-based Quasicontinuum Approximation

Mitchell Luskin, Christoph Ortner|Warwick Research Archive Portal (University of Warwick)|Apr 10, 2011

Nonlocal and gradient elasticity in micro/nano structuresMaterials Science29 references3 citations

TL;DR

This paper analyzes linear stationary iterative methods for the force-based quasicontinuum (QCF) approximation in a one-dimensional model, focusing on stability and convergence. It rigorously establishes that the Richardson iteration and preconditioned methods with $ L^{\text{qcl}}_F $ achieve convergence rates independent of system size $ N $, while the ghost force correction (GFC) method fails to contract in any norm below the critical strain, revealing fundamental limitations in its reliability for certain regimes.

ABSTRACT

Force-based multiphysics coupling methods have become popular since they provide a simple and efficient coupling mechanism, avoiding the difficulties in formulating and implementing a consistent coupling energy. They are also the only known pointwise consistent methods for coupling a general atomistic model to a finite element continuum model. However, the development of efficient and reliable iterative solution methods for the force-based approximation presents a challenge due to the non-symmetric and indefinite structure of the linearized force-based quasicontinuum approximation, as well as to its unusual stability properties. In this paper, we present rigorous numerical analysis and computational experiments to systematically study the stability and convergence rate for a variety of linear stationary iterative methods.

Motivation & Objective

• To address the challenge of developing efficient and reliable iterative solvers for the non-symmetric, indefinite linear system arising from the force-based quasicontinuum (QCF) approximation.
• To analyze the stability and convergence behavior of linear stationary iterative methods, particularly in the presence of unusual stability properties like indefiniteness in the QCF formulation.
• To evaluate the performance of key iterative methods—Richardson, preconditioned with $ L^{\text{qcl}}_F $, and ghost force correction (GFC)—in terms of contraction and convergence rates.
• To determine under what conditions these methods remain stable and converge, especially near the critical strain $ F_* $, where the system becomes ill-conditioned.
• To provide rigorous bounds on the convergence rate and operator norms for iterative schemes in discrete Sobolev spaces.

Proposed method

• The study focuses on the linearized QCF system about a uniform deformation $ y^F $, modeling the iterative update as $ P(u^{(n+1)} - u^{(n)}) = \alpha r^{(n)} $, where $ P $ is a preconditioner and $ \alpha > 0 $ is a fixed damping parameter.
• The Richardson iteration is analyzed with $ P = I $, and a contraction rate of order $ 1 - O(N^{-2}) $ is derived in the $ \ell^p_\varepsilon $ norm, showing convergence that improves with system size.
• An elliptic preconditioner $ P = L^{\text{qcl}}_F $, a standard second-order elliptic operator, is used to precondition the system, and convergence is proven in the $ \mathcal{U}^{2,\infty} $ norm up to the critical strain $ F_* $ with an optimal damping parameter $ \alpha $.
• The ghost force correction (GFC) iteration uses $ P = L^{\text{qce}}_F $, and it is shown that this method fails to be a contraction in any discrete Sobolev norm below $ F_* $, indicating potential instability.
• Operator norms of the iteration matrices are computed using conjugate operators and generalized eigenvalue problems, with bounds derived for $ \|G_{\text{qce}}\|_{\mathcal{U}^{1,p}} $ up to a factor of 2 for $ p = 1, \infty $.
• Theoretical analysis is complemented by computational experiments to validate convergence behavior and identify critical strain thresholds.

Experimental results

Research questions

• RQ1 Does the Richardson iteration converge for the linearized force-based quasicontinuum system, and what is its convergence rate in terms of system size $ N $?
• RQ2 Can preconditioning with the $ L^{\text{qcl}}_F $ operator ensure convergence up to the critical strain $ F_* $, and is the convergence rate independent of $ N $?
• RQ3 Does the ghost force correction (GFC) iteration remain a contraction in any norm below the critical strain $ F_* $, indicating reliable convergence?
• RQ4 What are the precise operator norms of the iteration matrices in discrete Sobolev spaces, and how do they depend on the strain and damping parameter $ \alpha $?
• RQ5 How do the stability and convergence properties of iterative solvers differ between the QCF, QCE, and QCL formulations in the linearized regime?

Key findings

• The Richardson iteration achieves a contraction rate of $ 1 - O(N^{-2}) $ in the $ \ell^p_\varepsilon $ norm, indicating convergence that improves with increasing system size.
• Preconditioning with $ L^{\text{qcl}}_F $ yields a contraction in the $ \mathcal{U}^{2,\infty} $ norm up to the critical strain $ F_* $, with a convergence rate independent of $ N $ when the damping parameter $ \alpha $ is optimally chosen.
• The ghost force correction (GFC) iteration fails to be a contraction in any discrete Sobolev norm below the critical strain $ F_* $, indicating potential instability and unreliability in this regime.
• The optimal damping parameter for the $ L^{\text{qcl}}_F $-preconditioned iteration is derived explicitly as $ \alpha^{\text{qcl,1,} \infty}_{\text{opt}} $, ensuring minimal contraction rate.
• The operator norm $ \|G_{\text{qce}}\|_{\mathcal{U}^{1,p}} $ is bounded within a factor of 2 of the $ \ell^p_\varepsilon $-norm of the conjugate operator, enabling practical estimation of convergence behavior.
• The analysis reveals that the QCF system's indefiniteness and non-symmetric structure lead to fundamental challenges in iterative solution, with only carefully preconditioned methods achieving robust convergence.